Key points
- METR found Anthropic’s model capability improvements are occurring at twice the rate recorded one year earlier.
- CEO Dario Amodei publicly urged the industry to slow AI capability gains to allow safety measures to keep pace.
- The assessment occurred as Anthropic prepares for a potential record-breaking initial public offering later this month.
Anthropic is accelerating the development of its artificial intelligence systems at approximately double the speed observed twelve months prior, according to an analysis released by an independent monitoring group.reported This finding emerges as the company’s chief executive has publicly advocated for a deceleration in the industry’s improvement pace to ensure safety mechanisms can adequately address emerging risks.This finding emerges as the company’s chief executive has publicly advocated for a deceleration in the industry’s improvement pace to ensure safety mechanisms can adequately address emerging risks.This finding emerges as the company’s chief executive has publicly advocated for a deceleration in the industry’s improvement pace to ensure safety mechanisms can adequately address emerging risks.
Evidence: reported
Anthropic’s AI development has doubled in speed despite slowdown calls, research group saysaifleet-fastcompany
The evidence
Researchers from Model Evaluation and Threat Research, a Berkeley-based nonprofit, conducted the assessment by reviewing internal data and interviewing staff at the San Francisco-based firm. The group estimated that Anthropic is enhancing its models’ performance on well-specified tasks around twice as fast as it was a year ago. Much of this accelerated development appears driven by the use of AI itself to aid in the improvement process.
The decision
During the period of METR’s assessment, Anthropic CEO Dario Amodei published an essay on 12 September responding to concerns about runaway AI.reported He urged the broader AI community to slow the rate at which they improve model capabilities, arguing that safety measures need time to catch up with technical advancements. The company did not immediately provide comment on the report when contacted for response.
Evidence: reported
Anthropic’s AI development has doubled in speed despite slowdown calls, research group saysaifleet-fastcompany
What the evidence means
The findings highlight a tension between corporate leadership rhetoric and internal development velocity as Anthropic prepares for its Wall Street debut. Details about the technology and its associated risks have primarily come from AI companies themselves, which have been encouraged by the Trump administration to self-police on safety matters. The potential for AI models to elude human control with catastrophic consequences remains a central debate.
What builders can do with it
Developers can observe that Anthropic’s internal processes increasingly rely on AI to accelerate capability gains for well-specified tasks. Builders should note that the company’s development velocity has doubled in the past year, suggesting that automated improvement loops are becoming a significant factor in model iteration. However, the documented limits of these capabilities remain opaque, as the company has not publicly detailed specific technical constraints or failure modes in this context.
Why it matters
- Policy makers must consider that self-regulation may not align with internal development incentives when companies accelerate capability gains while publicly calling for caution.
- Investors evaluating Anthropic’s upcoming public offering should weigh the risk of rapid capability escalation against the company’s stated safety commitments.
- Researchers can use the METR findings as a baseline for understanding how AI-assisted development impacts the velocity of model improvement in leading firms.
- Safety engineers may need to adapt their evaluation methods to keep pace with development cycles that are doubling in speed.
Topics: Anthropic, METR, AI safety, IPO, model development
Updated
The strongest doubt
Counterpoint from the sources
The company did not immediately comment on the report, leaving the specific internal drivers for the accelerated pace unconfirmed by Anthropic itself.
For readers
Policy
- Self-regulation may be insufficient if development velocity outpaces safety rhetoric.
- The Trump administration encourages companies to police themselves on AI safety matters.
Investors
- Anthropic plans to go public later this month in a potentially record-breaking IPO.
- Rapid capability gains may introduce unforeseen risks for public shareholders.
Researchers
- METR found that AI is playing a bigger role in its own improvement process.
- Independent monitoring groups can access internal data through invited assessments.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Terms in this story
METR
Model Evaluation and Threat Research, a Berkeley-based nonprofit that monitors AI safety and development.
Well-specified tasks
Clearly defined objectives or benchmarks used to measure AI model performance improvements.
Self-police
The practice of companies regulating their own safety standards without external government enforcement.
Questions readers ask
How much faster is Anthropic developing AI now compared to last year?
An independent analysis by METR estimates that Anthropic is improving its AI capabilities at around twice the rate it was a year ago.
Did Anthropic comment on the METR findings?
Anthropic did not immediately comment on the report in response to inquiries from reporters.
Why is this development pace significant for safety?
The rapid acceleration occurs despite CEO Dario Amodei’s public calls for the industry to slow down so that safety measures can catch up with capability improvements.
Connected stories
Reading receipt
- METR found Anthropic’s model capability improvements are occurring at twice the rate recorded one year earlier.
- CEO Dario Amodei publicly urged the industry to slow AI capability gains to allow safety measures to keep pace.
- The assessment occurred as Anthropic prepares for a potential record-breaking initial public offering later this month.
Still open: How exactly is Anthropic using AI to accelerate its own model development?
Still open: What specific safety measures are lagging behind the doubled pace of capability improvements?